Firmus IPO, September 2026: Goodman's Valuation Draws the 'Heroic' Label as 900 MW Pipeline Meets Scepticism
Firmus has launched a global IPO roadshow backed by a Goodman Group valuation that investors have called heroic, set against a contracted capacity pipeline reportedly exceeding 900 MW and a reported $2 billion August funding round at over $10.5 billion.
Firmus has taken its case to global institutional investors this month, launching an ASX IPO roadshow alongside a valuation supplied by Goodman Group that some market participants have described as "heroic." The phrase, as reported by The Australian, captures an investor divide that goes well beyond one company's listing: it is a test of whether AI infrastructure assets at the development stage — however large their contracted pipeline — can command multiples that listed data centre operators and neoclouds have not yet established as sustainable.
The company's trajectory has been rapid by any measure. Firmus began as a Bitcoin mining operator before pivoting to AI infrastructure, compressing what would typically be years of capital formation into a single year. In April, the company raised $505 million with Nvidia participating as a backer. By August, it had secured a further reported $2 billion from Coatue, Nvidia, Blackstone, and Jane Street, at a valuation cited by multiple sources at over $10.5 billion — roughly double the figure implied by the April round. One outlet, SmartCompany, reported the round at $2.85 billion and a $15 billion valuation; the higher figures have not been confirmed across the broader coverage, and the spread itself — unusual for a round of this scale — suggests the transaction structure may contain tranches, conditions, or warrants that complicate the headline number.
The commercial pipeline assembled in support of the IPO is substantial. Firmus and Nvidia announced a compute partnership in Indonesia — a 170,000-GPU campus in Batam described at announcement as the largest project of its kind. A multi-year compute agreement with OpenAI in Malaysia, framed as an extension of the Stargate programme into Southeast Asia, pushed total contracted capacity past 900 MW, according to Pulse 2.0. In Australia, a $300 million partnership with Slattery targets domestic AI-dedicated data centre development, and a third campus in Wesley Vale, Tasmania — targeting 52 MW — has been submitted for planning approval. Infrastructure contractor Maas Group separately disclosed an A$855 million contract for data centre and power work tied to the Firmus programme, providing an independently confirmed data point for the construction programme's scale that does not rely on Firmus's own disclosures.
The bull case for the valuation rests on Firmus's structural position inside Nvidia's commercial ecosystem, its role in the Stargate geographic expansion, and a degree of counterparty and geographic diversification that few comparable operators at this stage have achieved. Neocloud businesses anchored to a single hyperscaler or a single market have historically been vulnerable when demand timing slips; Firmus's spread across Australia, Indonesia, and Malaysia — with OpenAI, Nvidia, and Goodman serving as distinct anchor relationships — provides some structural buffer. A connectivity and grid-upgrade agreement with SUBCO, announced this month, suggests the company is also attempting to pre-solve the power constraints that have limited peers in competing markets.
The scepticism is equally grounded. The "heroic" characterisation reflects a view, as The Australian reported, that the IPO embeds utilisation, cost-of-capital, and construction-timeline assumptions that a company still assembling its first facilities cannot yet validate. Firmus's origins in Bitcoin mining remain a scrutiny point: the operational culture, customer relationships, and engineering disciplines of a cryptocurrency mining business differ substantially from those required to run hyperscaler-grade colocation at scale. The pipeline figures themselves carry a degree of promotional risk inherent in pre-construction announcements: the $30 billion Indonesia compute partnership has been announced but not independently verified at a component level, and contracted capacity figures — however large — depend on counterparties actually drawing down commitments on schedule.
Three concrete signals will determine whether the investor divide resolves in Firmus's favour. The first is IPO book quality: whether the listing prices at or near the Goodman-derived valuation with institutional demand to match, or whether the book requires material concessions. The second is first-megawatt delivery: commissioned and billed capacity in Indonesia and Malaysia will be the earliest hard test of whether execution tracks the pipeline announcements. The third is the Nvidia revenue-participation structure: separately reported, Nvidia has acted as a GPU backstop for neocloud customers — guaranteeing repurchase of idle capacity — in exchange for a share of inference-cloud revenue. The specific terms of that arrangement at Firmus will be material to any long-run margin forecast. Until those data points are public, the debate over what Goodman's valuation actually implies will not resolve.